[LoongArch] Optimize inserting extracted elements

Author: zhaoqi5

llvm/lib/Target/LoongArch/LoongArchLASXInstrInfo.td

@@ -1593,11 +1593,14 @@ def : Pat<(vector_insert v8i32:$xd, GRLenVT:$rj, uimm3:$imm),
           (XVINSGR2VR_W v8i32:$xd, GRLenVT:$rj, uimm3:$imm)>;
 def : Pat<(vector_insert v4i64:$xd, GRLenVT:$rj, uimm2:$imm),
           (XVINSGR2VR_D v4i64:$xd, GRLenVT:$rj, uimm2:$imm)>;
-
-def : Pat<(vector_insert v8f32:$vd, FPR32:$fj, uimm3:$imm),
-          (XVINSGR2VR_W $vd, (COPY_TO_REGCLASS FPR32:$fj, GPR), uimm3:$imm)>;
-def : Pat<(vector_insert v4f64:$vd, FPR64:$fj, uimm2:$imm),
-          (XVINSGR2VR_D $vd, (COPY_TO_REGCLASS FPR64:$fj, GPR), uimm2:$imm)>;
+def : Pat<(vector_insert v8f32:$xd, (f32 (vector_extract v8f32:$xj, uimm3:$imm1)), uimm3:$imm2),
+          (XVINSGR2VR_W $xd, (XVPICKVE2GR_W v8f32:$xj, uimm3:$imm1), uimm3:$imm2)>;
+def : Pat<(vector_insert v4f64:$xd, (f64 (vector_extract v4f64:$xj, uimm2:$imm1)), uimm2:$imm2),
+          (XVINSGR2VR_D $xd, (XVPICKVE2GR_D v4f64:$xj, uimm2:$imm1), uimm2:$imm2)>;
+def : Pat<(vector_insert v8f32:$xd, FPR32:$fj, uimm3:$imm),
+          (XVINSGR2VR_W $xd, (COPY_TO_REGCLASS FPR32:$fj, GPR), uimm3:$imm)>;
+def : Pat<(vector_insert v4f64:$xd, FPR64:$fj, uimm2:$imm),
+          (XVINSGR2VR_D $xd, (COPY_TO_REGCLASS FPR64:$fj, GPR), uimm2:$imm)>;
 
 // scalar_to_vector
 def : Pat<(v8f32 (scalar_to_vector FPR32:$fj)),

llvm/lib/Target/LoongArch/LoongArchLSXInstrInfo.td

@@ -1791,7 +1791,10 @@ def : Pat<(vector_insert v4i32:$vd, GRLenVT:$rj, uimm2:$imm),
           (VINSGR2VR_W v4i32:$vd, GRLenVT:$rj, uimm2:$imm)>;
 def : Pat<(vector_insert v2i64:$vd, GRLenVT:$rj, uimm1:$imm),
           (VINSGR2VR_D v2i64:$vd, GRLenVT:$rj, uimm1:$imm)>;
-
+def : Pat<(vector_insert v4f32:$vd, (f32 (vector_extract v4f32:$vj, uimm2:$imm1)), uimm2:$imm2),
+          (VINSGR2VR_W $vd, (VPICKVE2GR_W v4f32:$vj, uimm2:$imm1), uimm2:$imm2)>;
+def : Pat<(vector_insert v2f64:$vd, (f64 (vector_extract v2f64:$vj, uimm1:$imm1)), uimm1:$imm2),
+          (VINSGR2VR_D $vd, (VPICKVE2GR_D v2f64:$vj, uimm1:$imm1), uimm1:$imm2)>;
 def : Pat<(vector_insert v4f32:$vd, FPR32:$fj, uimm2:$imm),
           (VINSGR2VR_W $vd, (COPY_TO_REGCLASS FPR32:$fj, GPR), uimm2:$imm)>;
 def : Pat<(vector_insert v2f64:$vd, FPR64:$fj, uimm1:$imm),

llvm/test/CodeGen/LoongArch/lasx/ir-instruction/fix-xvshuf.ll

@@ -7,20 +7,12 @@ define <4 x double> @shufflevector_v4f64(<4 x double> %a, <4 x double> %b) {
 ; CHECK-LABEL: shufflevector_v4f64:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    xvpickve2gr.d $a0, $xr0, 0
-; CHECK-NEXT:    movgr2fr.d $fa2, $a0
-; CHECK-NEXT:    xvpickve2gr.d $a0, $xr1, 2
-; CHECK-NEXT:    movgr2fr.d $fa3, $a0
-; CHECK-NEXT:    movfr2gr.d $a0, $fa2
 ; CHECK-NEXT:    xvinsgr2vr.d $xr2, $a0, 0
-; CHECK-NEXT:    movfr2gr.d $a0, $fa3
+; CHECK-NEXT:    xvpickve2gr.d $a0, $xr1, 2
 ; CHECK-NEXT:    xvinsgr2vr.d $xr2, $a0, 1
 ; CHECK-NEXT:    xvpickve2gr.d $a0, $xr0, 3
-; CHECK-NEXT:    movgr2fr.d $fa0, $a0
-; CHECK-NEXT:    xvpickve2gr.d $a0, $xr1, 3
-; CHECK-NEXT:    movgr2fr.d $fa1, $a0
-; CHECK-NEXT:    movfr2gr.d $a0, $fa0
 ; CHECK-NEXT:    xvinsgr2vr.d $xr2, $a0, 2
-; CHECK-NEXT:    movfr2gr.d $a0, $fa1
+; CHECK-NEXT:    xvpickve2gr.d $a0, $xr1, 3
 ; CHECK-NEXT:    xvinsgr2vr.d $xr2, $a0, 3
 ; CHECK-NEXT:    xvori.b $xr0, $xr2, 0
 ; CHECK-NEXT:    ret

llvm/test/CodeGen/LoongArch/lasx/ir-instruction/insert-extract-element.ll

@@ -61,8 +61,6 @@ define <8 x float> @insert_extract_v8f32(<8 x float> %a) nounwind {
 ; CHECK-LABEL: insert_extract_v8f32:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    xvpickve2gr.w $a0, $xr0, 7
-; CHECK-NEXT:    movgr2fr.w $fa1, $a0
-; CHECK-NEXT:    movfr2gr.s $a0, $fa1
 ; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a0, 1
 ; CHECK-NEXT:    ret
 entry:
@@ -87,8 +85,6 @@ define <4 x double> @insert_extract_v4f64(<4 x double> %a) nounwind {
 ; CHECK-LABEL: insert_extract_v4f64:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    xvpickve2gr.d $a0, $xr0, 3
-; CHECK-NEXT:    movgr2fr.d $fa1, $a0
-; CHECK-NEXT:    movfr2gr.d $a0, $fa1
 ; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a0, 1
 ; CHECK-NEXT:    ret
 entry:

llvm/test/CodeGen/LoongArch/lsx/ir-instruction/insert-extract-element.ll

@@ -40,8 +40,7 @@ entry:
 define <4 x float> @insert_extract_v4f32(<4 x float> %a) nounwind {
 ; CHECK-LABEL: insert_extract_v4f32:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vreplvei.w $vr1, $vr0, 3
-; CHECK-NEXT:    movfr2gr.s $a0, $fa1
+; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 3
 ; CHECK-NEXT:    vinsgr2vr.w $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
@@ -65,8 +64,7 @@ entry:
 define <2 x double> @insert_extract_v2f64(<2 x double> %a) nounwind {
 ; CHECK-LABEL: insert_extract_v2f64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vreplvei.d $vr1, $vr0, 1
-; CHECK-NEXT:    movfr2gr.d $a0, $fa1
+; CHECK-NEXT:    vpickve2gr.d $a0, $vr0, 1
 ; CHECK-NEXT:    vinsgr2vr.d $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry: